CN107908722B - Reverse k ranking query method based on distance - Google Patents
Reverse k ranking query method based on distance Download PDFInfo
- Publication number
- CN107908722B CN107908722B CN201711120601.6A CN201711120601A CN107908722B CN 107908722 B CN107908722 B CN 107908722B CN 201711120601 A CN201711120601 A CN 201711120601A CN 107908722 B CN107908722 B CN 107908722B
- Authority
- CN
- China
- Prior art keywords
- merchant
- query
- grid
- mobile user
- index
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/24—Querying
- G06F16/245—Query processing
- G06F16/2453—Query optimisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/22—Indexing; Data structures therefor; Storage structures
- G06F16/2228—Indexing structures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/24—Querying
- G06F16/242—Query formulation
- G06F16/2425—Iterative querying; Query formulation based on the results of a preceding query
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
Abstract
The invention discloses a reverse k ranking query method based on distance, which comprises the steps of firstly, using a grid index to index the current positions of a merchant and a mobile user, using a KD tree to index all dimensions of the merchant, and using an equal-width histogram to index user preference; then based on the three indexes, using a method based on lower boundary pruning, k users most interested in a specified business are quickly found. The method overcomes the natural defects in reverse ranking query based on distance, can return k potential customers for any merchant, is suitable for potential customer orientation with the merchant as the center in a mobile environment, and provides a solution for personalized precise marketing in location-based service.
Description
Technical Field
The invention belongs to the field of reverse ranking query in database technology, and particularly relates to a potential customer searching method taking a merchant as a center in a mobile environment. The reverse k ranking query method based on the distance mainly solves the problem of how to quickly return k potential customers most interested in a merchant q given to query the merchant q in a mobile environment.
Background
With the rapid development of wireless positioning technology and the popularization of various mobile devices, and the wide use of Web 2.0 applications, merchants can easily collect interest and preference of users, and can push some promotion information to some potential customers according to the current position and preference of mobile users in real time. The traditional personalized recommendation method based on machine learning has the advantages of long model training time and high algorithm complexity, and is difficult to adapt to the real-time requirement in a mobile environment. The existing reverse ranking query (reverse top-k query) based on distance can only find potential customers for some popular merchants, and for the merchants which are not popular, because the k value is unpredictable in advance, the returned result set of the query is often empty, that is, suitable potential customers cannot be found for the non-popular merchants.
Disclosure of Invention
The invention aims to find k customers most interested in a given inquiry merchant in real time in a mobile environment. In order to ensure the real-time performance of the query response, the invention establishes corresponding indexes for the merchant set P, the mobile user preference W, the merchant position and the current position of the mobile user offline. In the online query stage, according to the query merchant q, various indexes established offline are combined, and some unnecessary calculations are filtered out, so that the query efficiency is improved, and the query real-time performance under the mobile environment is met.
The specific technical scheme for realizing the invention is as follows:
a reverse k ranking query method based on distance firstly imports a data set which accords with the specification and comprises a merchant set P, a mobile user set M and a mobile user preference set W. Then, establishing a KD tree index for a merchant data set in an off-line manner, establishing an equal-width histogram for a mobile user preference set W, and establishing a grid index for the merchant position and the current position of the mobile user; and finally, determining the grid C where the query merchant is located, traversing the neighbor grid C 'of the grid C in a width-first mode, and performing pruning judgment by taking a histogram bucket in the grid C' as a basic unit, thereby avoiding some unnecessary calculations. The method specifically comprises the following steps:
step 1: inputting a merchant set P, a mobile user set M and a mobile user preference set W;
step 2: indexes are established on the sets P, M and W to respectively obtain a KD tree index, an equal-width histogram index and a grid index G;
and step 3: setting query parameters q and k, wherein q represents a query merchant, and k represents the number of mobile users needing to be returned;
a query merchant q initiates a query request and requires to return k mobile users most interested in q;
and 4, step 4: and positioning the grid C where the query merchant is located on the grid index G, searching the neighbor grid C 'of the grid C by using width-first search, judging whether the histogram barrel of the mobile user in the grid C' can be pruned by using an algorithm Filterbucket, if so, accessing the next histogram barrel, otherwise, calculating the accurate ranking value of the mobile user in the histogram barrel, and finally returning k mobile users with the top ranking values as query results.
The step 1 specifically comprises:
a1: inputting a merchant set P, wherein each merchant is described as a d + 1-dimensional vector P, wherein the d-th dimension and the d + 1-th dimension are the position attributes of the current merchant, namely the space longitude and latitude, and the other dimensions are the descriptions of the non-position attributes of the merchant;
a2: inputting a mobile user set M, and recording the current position of each mobile user;
a3: inputting a set W of interest preferences of mobile users, the preference of each mobile user m is described as a d-dimensional weight vector mw,mw (i)Represents a degree of preference of the user m for the ith attribute of the merchant, andwherein m isw (d)Representing user m's preference in the distance dimension, and the other dimensions are preferences in the non-distance dimension.
The step 2 specifically comprises:
b1: establishing KD tree indexes for all attributes on the merchant set P;
b2: establishing an equal-width histogram index for the mobile user preference set W;
b3: and establishing a grid index G for the position attribute of the merchant set P and the current position of the mobile user.
The step 4 specifically includes:
c1: finding the position of a grid C where the query commercial tenant q is located on the grid index G;
c2: accessing neighbor nodes of C one by one in a width-first order, namely starting from a one-hop neighbor of the grid C, performing width-first search, and accessing a neighbor grid C' of the C which is not accessed yet;
c3: searching leaf nodes and brother nodes thereof in the KD tree, which are coincident with the C' position, searching n ancestor nodes of the leaf nodes and the brother nodes, forming a plurality of subtrees, and performing the step C4 on each subtree nroot one by one;
c4: for contour histogram bucket L [ i ] that has not been visited in C]Based on nroot, the histogram bucket L [ i ] is judged using the algorithm FilterBucket]If the L [ i ] can be pruned safely, if the L [ i ] can be pruned, the step C4 is repeated to continuously judge the next L [ i ]]. If histogram bucket L [ i ]]Can not be safely pruned, returning a pending KD tree node queue Q according to an algorithm FilterBucketL[i]Calculating L [ i ]]The precise ranking of each mobile user to the query merchant q;
c5: according to the accurate rank of the query merchant q, selecting k users with the top rank to form a query return result set Z;
c6: returning to the step C4, continuing to judge the next histogram equal height bucket L [ i ] in C 'until all histogram equal height buckets in C' are accessed;
c7: and returning to the steps C2-C6 until all the neighbor grids of the C are traversed.
The FilterBucket algorithm in the step C4 specifically includes:
d1: pressing nroot into queue Q, based on a scoring functionWhere dist (m, p) represents the distance between mobile user m and merchant p, the element at the head of queue Q is calculatedThe relation between the minimum circumscribed rectangle r and the query merchant q;
d2: if belowP (r, L [ i ]]) I.e. if and only ifIf true; then, L [ i ]]Is ranked by the lower bound value, i.e. L [ i ]]Lb, increasing the number of data points in r, i.e. | r |, i.e. calculating L [ i |)].Lb=L[i]Lb + | r |; judgment of L [ i]Lb, if greater than the worst rank of the k mobile users with the best rank so far, represented by min rank, then represents L i]Can be safely pruned, then continue to judge the next L [ i ]];
D3 if WtihinP (r, L [ i ]), that is, if and only if
If yes, the minimum circumscribed rectangle r on the current node is pressed into the pending KD tree node queue QL[i];
D4: if belowP (r, L [ i ]]) And WtihinP (r, L [ i ]]) If the sub-tree is not found to be the root, the sub-tree of the nroot is pressed into the queue Q, the step D2 is returned to continue judging until the queue Q is empty, and Q is obtainedL[i]。
The present invention overcomes the natural deficiencies in distance-based reverse ranking queries, and proposes a distance-based reverse k-rank query that can return k potential mobile users for any merchant. In addition, in order to meet the real-time requirement of the mobile environment on query response, the invention also discloses a method for responding the reverse k ranking query based on the distance. According to the method, various off-line established index structures including grid indexes, KD tree indexes and the like are effectively used, and fast pruning is realized on the basis of the index structures, so that the search space of the algorithm is sharply reduced, and the real-time performance of query response in a mobile environment is ensured.
Drawings
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a flow chart of algorithm 2Filter bucket of the present invention;
FIG. 3 is a schematic diagram of a breadth-first traversal of a grid over a grid index in accordance with the present invention;
FIG. 4 is a schematic diagram of the determination of the nroot subtree of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the following specific examples and drawings, and the present invention is not limited to the following examples. Variations and advantages that may occur to those skilled in the art may be incorporated into the invention without departing from the spirit and scope of the inventive concept, and the scope of the appended claims is intended to be protected.
The specific implementation mode of the invention is divided into two stages, wherein the first stage is the establishment of off-line index, and different index structures are established aiming at the access characteristics of multiple types of data objects for fast response to query. And in the second stage, unnecessary calculation is filtered based on the index established in the first stage, and the reverse k ranking query based on the distance is quickly responded, so that the real-time response requirement under the mobile environment is met.
Firstly, importing an external data set, including inputting a merchant set P, wherein each merchant is described as a d + 1-dimensional vector P, the d-th dimension and the d + 1-th dimension are the position attributes of the current merchant, namely the space longitude and latitude, and the other dimensions are the descriptions of the non-position attributes of the merchant; inputting a mobile user set M, and recording the current position of each mobile user; inputting a set W of interest preferences of mobile users, the preference of each mobile user m is described as a d-dimensional weight vector mw,mw (i)Represents a degree of preference of the user m for the ith attribute of the merchant, andwherein m isw (d)Representing user m's preference in the distance dimension, and the other dimensions are preferences in the non-distance dimension.
Then, respectively establishing indexes for the three imported data sets, including establishing KD tree indexes for all attributes on the merchant set P; establishing an equal-width histogram index for the mobile user preference set W; and establishing a grid index G for the position attribute of the merchant set P and the current position of the mobile user.
Finally, a query merchant q initiates a reverse k-rank query based on distance, and a specific processing flow of the query method based on reverse k-rank based on distance of the present invention is, as shown in fig. 1, first, according to a geographic location of the query merchant, on a grid index G, a grid location C (e.g., a solid grid in fig. 3) where the query merchant q is located is found, and neighbor grids of C are visited one by one in a width-first order, that is, starting from s-hop neighbors of the grid C (e.g., dashed lines in fig. 3 indicate 1-hop and 2-hop neighbors, and s ═ 0 indicates C itself), and a width-first search is performed to visit neighbor grids C' that have not been visited yet. Then, leaf nodes and sibling nodes thereof (as nodes framed by dotted lines in a KD tree schematic diagram in fig. 4, namely leaf nodes coincident with the C ' position) coincident with the C ' position in the KD tree are searched, respective N ancestor nodes are searched, the father node of the node is 1 ancestor, the grandfather node is 2 ancestor, N is a constant, and the part framed by a solid line in fig. 4 is 2 ancestor nodes, so that a plurality of subtrees are formed, whether the equal-height histogram buckets L [ i ] which are not accessed in the C ' can be safely pruned is judged one by one for each subtree, and the judgment is specifically finished by an algorithm FilterBucket and an intermediate result which cannot be definitely judged is returned. And if the histogram can be pruned, continuously executing the algorithm FilterBucket to judge the next unvisited histogram of equal height bucket L [ i ], otherwise, calculating the accurate ranking of each mobile user in the histogram of equal height bucket L [ i ] according to the result returned by the algorithm FilterBucket, updating the query return result set Z according to k users with the top ranking, and continuously traversing other histogram of equal height buckets which are unvisited in C 'until all the histogram of equal height buckets in C' are unvisited. At this point, the neighboring grid C' finishes processing, and continues to find a new neighboring grid of C that has not been visited yet, and all grids that know the grid index are visited and finished. And finally, returning the final query result set Z.
Examples
The present invention is named as algorithm 1 (responserrkranks), and in order to make the implementation flow more concise and clear, algorithm 1 includes algorithm 2 (FilterBucket).
Fig. 1 is a specific processing flow of the algorithm 1, and firstly inputs a merchant data set P, a mobile user data set M, and a mobile user preference data set W. Then, establishing indexes on each data set, including establishing grid indexes on the positions of merchants and mobile users; establishing KD tree indexes in a merchant full-dimensional space, and maintaining a minimum external rectangle which is the same as an R tree in each non-leaf node; a uniform width histogram H is built over the mobile user preferences.
After the corresponding index is established, the algorithm locates the grid C where the query merchant is located on the grid G, stores the query return results by using Z, namely k users with the best ranking currently, and records the accessed nodes in the KD tree by using a visited set in order to avoid repeated access of the nodes in the KD tree. With grid C as a central point, a breadth-first search is performed, where s-hop (s is 0,1, 2.) denotes a neighbor grid that is s hops away from grid C, and except for s being 0, there are generally 8 neighbor nodes with the same hop count. The outermost while loop in algorithm 1 represents a process, i.e., a process of traversing the mesh in the breadth-first manner.
Since all the mobile users' preferences are pre-divided into some effective buckets in the mode of contour histogram (according to whether the sum of preference vectors in the buckets meets the requirement or not)To determine the validity of the bucket). For mobile users located in the neighbor grid C ', they can be mapped into several valid histogram-of-equal-height buckets according to their preference vectors, where all valid histogram-of-equal-height buckets corresponding to mobile users in C' are loaded with a list L, and each bucket L [ i ] in L is traversed]The algorithm 2FilterBucket is used to check if the bucket can be pruned safely, and if so, continue to decide the next histogram bucket L i in C](ii) a If the KD tree nodes can not be pruned, the FilterBucket loads the KD tree nodes needing further accurate calculation and the minimum external rectangles on the KD tree nodes into a queue QL[i]Returning to algorithm 1. Algorithm 1 further calculates L [ i ]]For the accurate ranking value of the query merchant q, updating the query return result set Z according to k mobile users with the best ranking, and furtherA new threshold parameter min rank representing the worst ranking value of the k mobile users having the best ranking so far. After all the corresponding histogram buckets in C' are processed, the next neighbor node of the grid C where the merchant is located is searched and inquired continuously according to the width-first strategy until all the neighbor nodes are accessed.
The specific processing flow of algorithm 2(FilterBucket) is shown in fig. 2. Wherein the input parameters include: a KD tree sub-tree nroot coinciding with the neighbor grid C'; user preference equal-width histogram bucket L [ i ] of current visit located in C](ii) a At the current L [ i ]]Next, KD tree node queue Q which needs to return to main program for accurate calculationL[i](ii) a Querying a commercial tenant q; the ranking value min _ rank with the best ranking among the k users returned so far; and recording all already accessed KD tree node sets visited in order to avoid duplicate computations on the KD tree. The nroot is first pushed into queue Q based on a scoring functionWhere dist (m, p) represents the distance between mobile user m and merchant p. The relationship between the minimum bounding rectangles r and Q of the elements at the head of queue Q is calculated. If belowP (r, L [ i ]]) (if and only ifIs true, then L [ i]The lower bound value of (c) increases the number of data points in r, i.e., L [ i].Lb=L[i]Lb + | r |. Judgment of L [ i]Lb, if greater than the worst ranking of the k mobile users with the best ranking so far (min rank representation), then represents L i]Can be safely pruned, then returns to the algorithm ResponseRkranks to judge the next L [ i]. If WtihinP (r, L [ i ]) is present]) (if and only if,if yes, the minimum circumscribed rectangle r on the current node is pressed into the queue QL[i]. If belowP (r, L [ i ]]) And WtihinP (r, L [ i ]]) If the sub-tree is not found, the sub-tree of the nroot is pressed into the queue Q, and the relation between the minimum circumscribed rectangles r and Q of the elements at the head of the queue Q is continuously judged until the queue Q is empty. Finally, the process is carried out in a batch,will QL[i]And returning to an algorithm ResponseRkrankranks for further accurate ranking calculation.
The invention adopts a mode of establishing an index structure off-line and inquiring response on line in real time, and overcomes the defects of long model training time and poor real-time performance of the traditional machine learning method. The k potential customers of most interest to a given merchant can be quickly found for that merchant in a mobile environment. The query method has wide application prospect in the field of personalized accurate marketing based on the position.
Claims (2)
1. A reverse k ranking query method based on distance is characterized by comprising the following steps:
step 1: inputting a merchant set P, a mobile user set M and a mobile user preference set W;
step 2: indexes are established on the sets P, M and W to respectively obtain a KD tree index, an equal-width histogram index and a grid index G;
and step 3: setting query parameters q and k, wherein q represents a query merchant, and k represents the number of mobile users needing to be returned; a query merchant q initiates a query request and requires to return k mobile users most interested in q;
and 4, step 4: positioning a grid C where a query merchant is located on a grid index G, searching a neighbor grid C 'of the grid C by using width-first search, judging whether a histogram bucket of a mobile user in the grid C' can be pruned by using an algorithm Filterbucket, if so, accessing a next histogram bucket, otherwise, calculating an accurate ranking value of the mobile user in the histogram bucket, and finally returning k mobile users with the top ranking values as query results; wherein:
the step 2 specifically comprises:
b1: establishing KD tree indexes for all attributes on the merchant set P, and maintaining a minimum external rectangle like an R tree at a middle node of the KD tree;
b2: establishing an equal-width histogram index for the mobile user preference set W;
b3: establishing a grid index G for the position attribute of the merchant set P and the current position of the mobile user;
the step 4 specifically includes:
c1: finding the position of a grid C where the query commercial tenant q is located on the grid index G;
c2: accessing neighbor nodes of C one by one in a width-first order, namely starting from a one-hop neighbor of the grid C, performing width-first search, and accessing a neighbor grid C' of the C which is not accessed yet;
c3: searching leaf nodes and brother nodes thereof in the KD tree, which are coincident with the C' position, searching n ancestor nodes of the leaf nodes and the brother nodes, forming a plurality of subtrees, and performing the step C4 on each subtree nroot one by one;
c4: for contour histogram bucket L [ i ] that has not been visited in C]Based on nroot, the histogram bucket L [ i ] is judged using the algorithm FilterBucket]If the L [ i ] can be pruned safely, if the L [ i ] can be pruned, the step C4 is repeated to continuously judge the next L [ i ]](ii) a If histogram bucket L [ i ]]Can not be safely pruned, returning a pending KD tree node queue Q according to an algorithm FilterBucketL[i]Calculating L [ i ]]The precise ranking of each mobile user to the query merchant q;
c5: according to the accurate rank of the query merchant q, selecting k users with the top rank to form a query return result set Z;
c6: returning to the step C4, continuing to judge the next histogram equal height bucket L [ i ] in C 'until all histogram equal height buckets in C' are accessed;
c7: returning to the steps C2-C6 until all the neighbor grids of the C are traversed;
the FilterBucket algorithm in the step C4 specifically includes:
d1: pressing nroot into queue Q, based on a scoring functionDist (m, p) represents the distance between the mobile user m and the merchant p, and the relation between the minimum circumscribed rectangle r of the node element at the head of the queue Q and the query merchant Q is calculated;
d2: if belowP (r, L [ i ]]) I.e. if and only ifIf true; then, L [ i ]]Is ranked by the lower bound value, i.e. L [ i ]]Lb, increasing the number of data points in r, i.e. | r |, i.e. calculating L [ i |)].Lb=L[i]Lb + | r |; judgment of L [ i]Lb, if greater than the worst rank of the k mobile users with the best rank so far, represented by min rank, then represents L i]Can be safely pruned, then continue to judge the next L [ i ]];
D3 if WtihinP (r, L [ i ]]) I.e. if and only ifIf yes, the minimum circumscribed rectangle r on the current node is pressed into the pending KD tree node queue QL[i];
D4: if belowP (r, L [ i ]]) And WtihinP (r, L [ i ]]) If the result is not true, the subtree of nroot is pushed into the queue Q, the step D2 is returned to continue judging until the queue Q is empty, and the pending KD tree node queue Q is obtainedL[i]。
2. The method according to claim 1, wherein step 1 specifically comprises:
a1: inputting a merchant set P, wherein each merchant is described as a d + 1-dimensional vector P, wherein the d-th dimension and the d + 1-th dimension are the position attributes of the current merchant, namely the space longitude and latitude, and the other dimensions are the descriptions of the non-position attributes of the merchant;
a2: inputting a mobile user set M, and recording the current position of each mobile user;
a3: inputting a set W of interest preferences of mobile users, the preference of each mobile user m is described as a d-dimensional weight vector mw,mw (i)Represents a degree of preference of the user m for the ith attribute of the merchant, andwherein m isw (d)Representing user m's preference in the distance dimension, and the other dimensions are preferences in the non-distance dimension.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711120601.6A CN107908722B (en) | 2017-11-14 | 2017-11-14 | Reverse k ranking query method based on distance |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711120601.6A CN107908722B (en) | 2017-11-14 | 2017-11-14 | Reverse k ranking query method based on distance |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107908722A CN107908722A (en) | 2018-04-13 |
CN107908722B true CN107908722B (en) | 2021-10-12 |
Family
ID=61845131
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711120601.6A Active CN107908722B (en) | 2017-11-14 | 2017-11-14 | Reverse k ranking query method based on distance |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107908722B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109657022B (en) * | 2018-12-08 | 2020-06-30 | 拉扎斯网络科技(上海)有限公司 | Merchant searching method and device, electronic equipment and storage medium |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101510222A (en) * | 2009-02-20 | 2009-08-19 | 北京大学 | Multilayer index voice document searching method and system thereof |
CN103778196A (en) * | 2014-01-07 | 2014-05-07 | 浙江大学 | Optical location method based on double-color reverse nearest neighbor query |
CN105378764A (en) * | 2013-07-12 | 2016-03-02 | 微软技术许可有限责任公司 | Interactive concept editing in computer-human interactive learning |
CN106503196A (en) * | 2016-10-26 | 2017-03-15 | 云南大学 | The structure and querying method of extensible storage index structure in cloud environment |
CN107145526A (en) * | 2017-04-14 | 2017-09-08 | 浙江大学 | Geographical social activity keyword Reverse nearest neighbor inquiry processing method under a kind of road network |
Family Cites Families (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20060212429A1 (en) * | 2005-03-17 | 2006-09-21 | Microsoft Corporation | Answering top-K selection queries in a relational engine |
US20120254153A1 (en) * | 2011-03-31 | 2012-10-04 | Microsoft Corporation | Shortest path determination in databases |
US20140337472A1 (en) * | 2012-12-13 | 2014-11-13 | Level 3 Communications, Llc | Beacon Services in a Content Delivery Framework |
US20150120717A1 (en) * | 2013-10-25 | 2015-04-30 | Marketwire L.P. | Systems and methods for determining influencers in a social data network and ranking data objects based on influencers |
-
2017
- 2017-11-14 CN CN201711120601.6A patent/CN107908722B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101510222A (en) * | 2009-02-20 | 2009-08-19 | 北京大学 | Multilayer index voice document searching method and system thereof |
CN105378764A (en) * | 2013-07-12 | 2016-03-02 | 微软技术许可有限责任公司 | Interactive concept editing in computer-human interactive learning |
CN103778196A (en) * | 2014-01-07 | 2014-05-07 | 浙江大学 | Optical location method based on double-color reverse nearest neighbor query |
CN106503196A (en) * | 2016-10-26 | 2017-03-15 | 云南大学 | The structure and querying method of extensible storage index structure in cloud environment |
CN107145526A (en) * | 2017-04-14 | 2017-09-08 | 浙江大学 | Geographical social activity keyword Reverse nearest neighbor inquiry processing method under a kind of road network |
Non-Patent Citations (5)
Title |
---|
efficient processing of exact top-k queries over disk-resident sorted lists;HweeHwa Pang等;《the VLDB journal》;20100630;第19卷(第3期);437-456页 * |
reverse k-ranks query;zhang zhao等;《proceedings of VLDB Endowment》;20140601;第7卷(第10期);785-796页 * |
一种基于逆支配点集的数据流Top-k计算方法;甘亮等;《计算机工程与科学》;20120615;第34卷(第6期);59-64页 * |
在线论坛用户兴趣图谱发现与个性化信息推荐;张召;《中国博士学位论文全文数据库》;20130715(第7期);I139-14页 * |
逆向序敏感查询在互联网广告投放中的应用;张召等;《华东师范大学学报(自然科学版)》;20130525(第3期);26-36+53页 * |
Also Published As
Publication number | Publication date |
---|---|
CN107908722A (en) | 2018-04-13 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105183921B (en) | The shop addressing system based on double-color reverse NN Query under mobile cloud computing environment | |
CN106960044B (en) | Time perception personalized POI recommendation method based on tensor decomposition and weighted HITS | |
CN109255054B (en) | Relation weight-based community discovery method in enterprise map | |
CN102521364B (en) | Method for inquiring shortest path between two points on map | |
CN107145526B (en) | Reverse-nearest neighbor query processing method for geographic social keywords under road network | |
JP2016503937A (en) | Search and determine activity areas | |
JP7407209B2 (en) | Information push method and device | |
CN104462263B (en) | A method of utilizing database index search shops | |
CN103995859B (en) | A kind of hot spot region incident detection system based on geographical labels applied to LBSN networks | |
CN102156756A (en) | Method for finding optimal path in road network based on graph embedding | |
CN104820715A (en) | Method and system for sharing and analyzing data based on multidimensional correlation | |
CN108932347A (en) | A kind of spatial key querying method based on society's perception under distributed environment | |
CN107220312A (en) | A kind of point of interest based on co-occurrence figure recommends method and system | |
CN105760469A (en) | High-dimensional approximate image retrieval method based on inverted LSH in cloud computing environment | |
CN104778237A (en) | Individual recommending method and system based on key users | |
CN108733803A (en) | A kind of Multi-User Dimension keyword query method under road network | |
CN105550332A (en) | Dual-layer index structure based origin graph query method | |
CN103778196B (en) | Optical location method based on double-color reverse nearest neighbor query | |
US20220027427A1 (en) | Business searching methods and apparatuses, electronic devices and storage media | |
CN104731887B (en) | A kind of user method for measuring similarity in collaborative filtering | |
CN107908722B (en) | Reverse k ranking query method based on distance | |
CN110148032A (en) | Products Show method, apparatus, storage medium and server based on geographical location | |
CN105138536B (en) | Mobile social networking data fragmentation method based on Directed Hypergraph | |
CN112507047B (en) | Optimal ordered path query method based on interest point preference | |
CN112069416B (en) | Cross-social network user identity recognition method based on community discovery |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |